NVIDIA Triton Inference Server provides a cloud and edge inferencing solution optimized for both CPUs and GPUs.

This top level GitHub organization host repositories for officially supported backends, including TensorRT, TensorFlow, PyTorch, Python, ONNX Runtime, and OpenVino. The organization also hosts several popular Triton tools, including:

• Model Analyzer: A tool to analyze the runtime performance of a model and provide an optimized model configuration for Triton Inference Server.

• Model Navigator: a tool that provides the ability to automate the process of moving a model from source to optimal format and configuration for deployment on Triton Inference Server.

To learn about NVIDIA Triton Inference Server, refer to the Triton developer page and read our Quickstart Guide. Official Triton Docker containers are available from NVIDIA NGC.

User documentation on Triton features, APIs, and architecture is located in the server documents on GitHub. A table of contents for the user documentation is located in the server README file.

Release Notes, Support Matrix, and Licenses information are available in the NVIDIA Triton Inference Server Documentation.

Specific end-to-end examples for popular models, such as ResNet, BERT, and DLRM are located in the NVIDIA Deep Learning Examples page on GitHub. Additional generic examples can be found in the server documents.

For technical questions about Triton Inference Server, please consult the Triton FAQ Guide. Information about future support & updates for Triton can be found in the Dynamo FAQ Guide.

Share feedback or ask questions about NVIDIA Triton Inference Server by filing a GitHub issue.